Capabilities
Everything YomiNinja Can Do for Your Japanese Immersion
From OCR engine selection to Anki mining — the complete feature set of the free, open-source Japanese game reader.
01 — Core
OCR text extraction
YomiNinja reads on-screen Japanese text using optical character recognition and makes it interactive. No game engine compatibility required — if text is visible as pixels, YomiNinja can read it.
Window or full-screen capture
Select any open application window as the capture source. YomiNinja identifies and tracks the window, even if it moves. Full-screen desktop capture is also available for edge cases.
OCR Templates — targeted capture regions
Draw and save a capture rectangle over just the dialogue box. YomiNinja crops every capture to this region before OCR processing, eliminating noise from UI elements and increasing both speed and accuracy. Templates are saved per game.
Auto OCR — hands-free continuous detection
Enable Auto OCR and put the hotkey away. YomiNinja monitors the capture region for changes using frame comparison. When new text appears — a new dialogue line, a menu update, a cutscene subtitle — OCR fires automatically within half a second, no interaction required.
Furigana threshold filtering
Furigana — small reading annotations above kanji — confuses OCR and breaks dictionary lookups. The furigana threshold slider (v0.9.1+) filters text below a configurable size, removing furigana from the OCR output while preserving the main kanji.
Image preprocessing
Adjustable contrast, brightness, and scale settings applied to the captured image before OCR processing. Preprocessing can significantly improve accuracy on low-contrast text, washed-out UI, or very small font sizes.
Five OCR engines to choose from
PaddleOCR (default, offline), MangaOCR (stylized fonts, offline), Google Cloud Vision (cloud, highest accuracy), Google Lens (cloud, no API key), and Apple Vision Framework (macOS, vertical text). Switch engines per game in one click.
Full engine comparison →02 — Dictionary
Built-in dictionary lookup
Yomitan and 10ten Reader ship pre-installed inside YomiNinja's Chromium context. No browser, no Web Store, no setup — open the app and hover any word.
Yomitan — pre-installed, no browser needed
Yomitan is the community standard for Japanese dictionary lookups. It supports JMdict, KANJIDIC, pitch accent dictionaries, frequency lists, and custom dictionary files. In YomiNinja, it works identically to how it does in Chrome — hover a word, popup appears with readings, definitions, and pitch.
10ten Reader — pre-installed alternative
10ten Reader is an alternative Japanese dictionary extension that also ships pre-installed. It uses a different UI layout and slightly different lookup behaviour — some learners prefer it. Switch between Yomitan and 10ten in YomiNinja's extension manager.
Inkah and additional Chrome extensions
YomiNinja includes a Chrome Extensions Manager. Install any Chromium-compatible extension — Inkah, Migaku, custom lookup tools — directly into YomiNinja's browser context without touching your main browser installation.
Dictionary setup guide →Browser overlay mode (PWA)
Alternative overlay mode renders the text layer inside a Progressive Web App in your default browser, rather than an Electron overlay window. This can offer better extension compatibility for configurations where the standard overlay has issues. Introduced in v0.8.3.
03 — Audio & Mining
Text-to-speech and Anki integration
VOICEVOX TTS — native Japanese speech
VOICEVOX is a high-quality, open-source Japanese text-to-speech engine with multiple natural-sounding voice characters. YomiNinja integrates VOICEVOX natively across all platforms (v0.9.1+): after OCR captures text, it can be read aloud automatically, reinforcing pronunciation patterns while you read.
WebSocket output — real-time text streaming
Every OCR result is broadcast over a local WebSocket server (default port 7331). Connect any application: a texthooker page, a logging script, a custom Anki pipeline, or a second-screen display. The WebSocket interface is always-on when YomiNinja is running.
Anki mining pipeline (via WebSocket)
Connect a texthooker page to YomiNinja's WebSocket feed. As you play, each captured sentence appears on the texthooker page. Hover any word with Yomitan, export to Anki. The complete loop from game dialogue to flashcard without leaving your desk setup.
Mining setup guide →04 — Platform
Performance, privacy, and platform support
Windows, Linux, macOS
Native packages for all three platforms: Windows installer (.exe), Linux AppImage/.deb/.rpm/.pacman, and macOS .dmg (Intel + Apple Silicon). Each platform has a tailored setup — see the download page for platform-specific requirements.
GPU acceleration — CUDA and ROCm
Reduce OCR latency from 200–400ms to under 100ms per capture. NVIDIA GPUs use CUDA (Windows + Linux, requires CUDA 11.8). AMD GPUs use ROCm (Linux only). GPU mode is most impactful in Auto OCR where captures fire continuously.
No data collection — fully private by default
YomiNinja collects zero telemetry, zero usage data, zero personal information. Local OCR engines (PaddleOCR, MangaOCR, Apple Vision) process everything on your machine. No account, no login, no cloud required unless you opt into a cloud OCR engine.
Fully open source — GPL-3.0
Every line of YomiNinja's code is publicly available on GitHub. Build from source, inspect the binary, contribute patches. The GPL-3.0 license guarantees it remains free and open.
Customizable hotkeys
Every YomiNinja action has a configurable keyboard shortcut: OCR trigger, overlay toggle, overlay move, Auto OCR on/off. Hotkeys are stored in settings and persist across sessions. Gamepad button mapping was also introduced in early versions.
Free forever — no subscription or tiers
Every feature on this page is available in the free download. There is no Pro tier, no premium subscription, no feature gating. Optional Patreon support gives early access to pre-release builds — nothing else is paywalled.
Get started
All features, free to use
Download YomiNinja and get the full feature set immediately — no account, no setup fees, no limitations.